Give non-multimodal models vision via a user-configured VLM API (OpenAI-compatible or Anthropic).
Updated 2026-08-10 10:28:50 +08:00